bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1 Spatial density estimates of Eurasian lynx (Lynx lynx) in the French Jura and Vosges
2 Mountains
3
4 Olivier Gimenez1, Sylvain Gatti2, Christophe Duchamp3, Estelle Germain4, Alain Laurent2,
5 Fridolin Zimmermann5, Eric Marboutin2
6
7 1CEFE, CNRS, Univ Montpellier, Univ Paul Valéry Montpellier 3, EPHE, IRD, Montpellier,
8 France
9 2Office National de la Chasse et de la Faune Sauvage, ZI Mayencin, Gières, France.
10 3Office National de la Chasse et de la Faune Sauvage, Parc Micropolis, Gap, France.
11 4Centre de Recherche et d’Observation sur les Carnivores (CROC), 4 rue de la banie, 57590,
12 Lucy, France
13 5KORA, Thunstrasse 31, 3074, Muri, Switzerland
14
15 Abstract
16
17 Obtaining estimates of animal population density is a key step in providing sound
18 conservation and management strategies for wildlife. For many large carnivores however,
19 estimating density is difficult because these species are elusive and wide-ranging. Here, we
20 focus on providing the first density estimates of the Eurasian lynx (Lynx lynx) in the French
21 Jura and Vosges mountains. We sampled a total of 413 camera trapping sites (with 2 cameras
22 per site) between January 2011 and April 2016 in seven study areas across seven counties of
23 the French Jura and Vosges mountains. We obtained 592 lynx detections over 19,035 trap
24 days in the Jura mountain and 0 detection over 6,804 trap days in the Vosges mountain. Based
25 on coat patterns, we identified a total number of 92 unique individuals from photographs, bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
26 including 16 females, 13 males and 63 individuals of unknown sex. Using spatial capture-
27 recapture (SCR) models, we estimated abundance in the study areas between 5 (SE = 0.1) and
28 29 (0.2) lynx and density between 0.24 (SE = 0.02) and 0.91 (SE = 0.03) lynx per 100 km2.
29 We also provide a comparison with non-spatial density estimates and discuss the expected
30 discrepancies. Our study is yet another example of the advantage of combining SCR methods
31 and non-invasive sampling techniques to estimate density for elusive and wide-ranging
32 species, like large carnivores. While the estimated densities in the French Jura mountain are
33 comparable to other lynx populations in Europe, the fact that we detected no lynx in the
34 Vosges mountain is alarming. Connectivity should be encouraged between the French Jura
35 mountain, the Vosges mountain and the Palatinate Forest in Germany where a reintroduction
36 program is currently ongoing. Our density estimates will help in setting a baseline
37 conservation status for the lynx population in France.
38
39 Introduction
40
41 Obtaining estimates of animal population density is a key step in providing sound
42 conservation and management strategies for wildlife [1]. For many large carnivores however,
43 estimating density is difficult because these species are elusive and wide-ranging, resulting in
44 low detection rates [2]. To deal with these issues, non-invasive techniques, such as camera
45 trapping and DNA sampling, are increasingly used [3]. These non-invasive techniques
46 generate data that can be analyzed with capture-recapture methods to estimate densities [4].
47 Standard capture-recapture models for closed populations [5] have long been used to
48 estimate animal abundance and density, including many large carnivores [6,7]. However,
49 when converting abundance into density, density estimates are highly sensitive to the size of
50 user-defined area assumed to reflect the effective sampling area [8]. In addition, individual bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
51 heterogeneity on the detection due to spatial variation in the distance of home ranges to the
52 sampling devices may lead to biased density estimates [5]. Spatial capture-recapture (SCR)
53 models deal with these issues by explicitly incorporating spatial locations of detections [9–
54 12], and they are increasingly used to estimate densities of large carnivores [13–18].
55 Here, we focus on the threatened Eurasian lynx (Lynx lynx) in the French Jura and
56 Vosges mountains (see [19] for a map of its distribution in Europe; see also
57 https://www.lcie.org/Large-carnivores/Eurasian-lynx for recent updates). As in many regions
58 of western Europe [20], lynx were extirpated from France between the 17th and 20th centuries
59 due to habitat degradation, persecution by humans and decrease in prey availability [21].
60 Shortly after their initial reintroduction in Switzerland in the 1970s [22], lynx naturally
61 increased their range and started recolonizing France by repopulating forests on the French
62 side of the Jura [21]. Reintroductions also occurred in the French Vosges mountain between
63 1983 and 1993 with the perspective of establishing a population there [23]. The species is
64 listed as endangered in the IUCN Red list and is of conservation concern in France due to
65 habitat fragmentation, poaching and collisions with cars and trains. Currently, the French
66 population of lynx is restricted to three mountain ranges, the Vosges, in northeastern France,
67 the Jura and the Alps, with little connectivity between them most likely due to human-made
68 linear infrastructures. While the Northern Alps are slowly being recolonized with lynx mostly
69 coming from the Jura [24], the Jura holds the bulk of the French lynx population. In contrast,
70 the lynx presence in the Vosges mountain remained stable following the reintroductions and
71 then, has been continuously decreasing since 2005 [25].
72 Despite their conservation status, little information on abundance and density of lynx
73 in France exist. In this study, we used SCR and standard capture-recapture models to provide
74 the first estimate of lynx abundance and density using camera-trap surveys implemented in bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
75 the French Jura and Vosges mountains from 2011 to 2016. Based on these results, we discuss
76 research and management priorities for the effective conservation of lynx in France.
77
78 Methods
79
80 Ethics statement
81 We used non-invasive methods for data collection, which did not involve manipulation or
82 handling of any living organism. Therefore, approval from an animal ethics committee was
83 not required. Cameras were set on public or private forests with the permission of local
84 authorities or local owners, respectively. We advertised the study and the presence of camera
85 traps to the local stakeholders and the public visiting the areas. In agreement with French
86 legislation, we deleted photos permitting the identification of goods or people.
87
88 Study area and sampling design
89 The study area encompassed three counties of the French Jura mountain, namely Ain, Doubs
90 and Jura and four counties of the Vosges mountain, namely Vosges, Haut-Rhin, Bas-Rhin and
91 Moselle (Figure 1). Elevation ranged from 163 and 1,718 m above sea level in the Jura
92 mountain and from 104 to 1,422 m in the Vosges mountain. The human population density
93 was 88 per km2 in the Jura mountain and 170 per km2 in the Vosges mountain. Forests cover
94 50% on average of the Jura mountain [26] and 70% of the Vosges mountain [27]. Sampling
95 occurred over 6 years, between January 2011 and April 2016, mostly in winter and spring,
96 with surveys lasting between 2 and 4 months. We considered two study areas in 2011, 2014
97 and 2015, three study areas in 2013 and one study area in 2012 and 2016 through camera
98 trapping (Figure 1).
99 bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
100 [Figure 1 about here]
101
102 We divided each study area into a grid of 2.7 x 2.7 km cells applying a systematic
103 design where one out of two cells was sampled [28], hence ensuring that at least one camera
104 trap was set in each potential lynx home range (between 100km2 and 250km2, see [29]). To
105 maximize detectability, we set (non-baited) camera traps in forested habitats, based on
106 previous signs of lynx presence and on local knowledge, at optimal locations where landscape
107 and terrain features were likely to channel lynx movements on more predictable paths (on
108 forest roads, hiking trails and to a lesser extent on game paths) [30]. Camera were settled on
109 within a single session design continuously during 60 days between February and beginning
110 of March with little variation between sites.
111 At each trapping location, we set two Xenon white flash camera traps (models:
112 Capture, Ambush and Attack; Cuddeback, WI, USA) with passive infrared trigger
113 mechanisms to photograph both flanks of an animal. We checked camera traps weekly to
114 change memory cards, batteries and to remove fresh snow after heavy snowfall. Based on
115 unique coat patterns, we identified individual lynx on photographs [31]. The recognition of
116 individual was computer-induced, not fully automated. We used the Extract-compare ©
117 software that compares the lynx spot pattern with a library of previously extracted pattern and
118 proposes potential matches according to a score
119 (http://conservationresearch.org.uk/Home/ExtractCompare). The observer can confirm the
120 lynx identification or not and browse through the highest-ranking proposed matches. The final
121 decision is made by the observer based on an additional visual examination of the entire photo
122 set for this particular lynx. Pictures for which no match was found with the software were
123 visually checked against our entire photo library. Only when the match is undeniable is the
124 individual recorded as a match, otherwise it was recorded as a new individual. All captures bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
125 that do not fit automated or associated visual confirmation with no doubt, because of a poor
126 picture quality (e.g. blurry, overexposed), were classified as “unconfirmed” and excluded
127 from the analyses. We recorded the date, time, sex whenever possible, and location of each
128 photographic capture of a lynx. During the time of year our study took place, juvenile lynx (<
129 1 year old) can still be with their mother [32]. In our analysis, we retained only independent
130 lynx, i.e. adult lynx or emancipated individuals based on physical characteristics or previous
131 knowledge of their age or status (from photographic evidence). We defined a capture occasion
132 as 5 successive trap nights [30], dissociating trapping events from individual photo to avoid
133 pseudo-replications.
134
135 Spatial capture-recapture analyses
136 We used spatial capture-recapture (SCR) models to estimate lynx densities [4]. In contrast
137 with standard (non-spatial) capture-recapture models, SCR models use the spatial locations of
138 captures to infer the activity center (or home range) of each individual. We assumed that
139 individual encounters are Bernoulli random variables with individual- and trap-specific
140 detection probabilities. More precisely, the detection probability pij of an individual i at trap j
141 is assumed to decrease as the distance (dij) from its activity center increases according to a
2 2 142 detection function. We used the half-normal detection function, pij = p0 exp(- dij /(2σ )),
143 where p0 is the probability of detecting an individual when the trap is located exactly at its
144 center of activity and σ is the spatial scale (or movement) parameter that controls the shape of
145 the detection function. For one of the two study areas in the French Jura mountain in years
146 2011 and 2013, we detected only a few individuals (see the columns Doubs in Table 1). To
147 increase the effective sample size, we combined the data from the two sampling areas using
148 common detection and spatial parameters for both areas, while estimating density separately
149 (e.g., [33]). We defined a state-space, i.e. the area encompassing all potential activity centers bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
150 of the observed individuals, by building a grid that buffered outermost camera trap locations
151 by 15 km (corresponding to at least 2σ [4]) with a resolution of 1.5 km (or pixels of area 2.25
152 km2). We fitted SCR models in the maximum likelihood framework using the R package
153 oSCR [34,35].
154 For comparison, we also estimated abundance using standard (non-spatial) capture-
155 recapture models [5]. We dropped the spatial information and considered only the detections
156 and non-detections for each individual. We considered two models, M0 in which the detection
157 probability is the same for all individuals, and Mh in which the detection probability varies
158 among individuals. We fitted standard models in the maximum likelihood framework using
159 the R package Rcapture [36]. We estimated density as the ratio of estimated abundance over
160 an effective trapping area (ETA). ETA was estimated by adding a buffer to the trapping area
161 equal to the mean maximum distance moved (MMDM) or half of it (HMMDM). We
162 calculated the MMDM by averaging the maximum distances between capture locations for all
163 individuals detected at more than one site.
164
165 Results
166
167 We collected data from 413 camera trapping sites (2 camera traps were set per site) resulting
168 in 25,839 trap days (Table 1). In total, we identified 92 lynx over 532 detection events in the
169 Jura mountain, including 16 females, 13 males and 63 individuals of unknown sex. The
170 number of detections per individual was 2.6 on average and varied from 1 up to 11. In
171 contrast, we collected no lynx photo in the Vosges mountain, therefore we did not proceed
172 with analyses for this area.
173
174 [Table 1 about here] bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
175
176 For the Jura mountain, abundance estimates were similar whether we used spatial or
177 non-spatial models, although always slightly higher for the former. Estimated abundance
178 among study areas varied between 5 (SE = 0.1) and 29 (0.2) lynx in the spatial analyses,
179 between 4 (0.7) and 23 (0.7) with model M0, and between 5 (1.7) and 28 (3.6) with model
180 Mh. Estimated density varied between 0.24 (0.02) and 0.91 (0.03) lynx per 100 km2 in the
181 spatial analyses (Table 2). In the non-spatial analyses, the density varied between 0.31 (0.05)
182 and 0.78 (0.02) lynx per 100 km2 under model M0 and between 0.34 (0.06) and 0.95 (0.12)
183 under model Mh when the MMDM was used. When we used HMMDM, the density varied
184 between 0.57 (0.10) and 1.46 (0.16) lynx per 100 km2 under model M0 and between 0.67
185 (0.12) and 1.43 (0.16) under model Mh.
186
187 [Table 2 about here]
188
189 From the spatial analyses, we used the model estimates to produce density surfaces
190 within the state-space (Figure 2). The density per pixel of area 2.25 km2 ranged from 0 to 0.20
191 individuals in the Jura mountain.
192
193 [Figure 2 about here]
194
195 Discussion
196
197 By using camera-trap sampling and SCR models, we provided the first multi-site density
198 estimates for lynx that will help in setting a baseline conservation status for the French lynx
199 population. The multi-site dimension of our study allows exploring variability in the density bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
200 estimates across landscapes. Our study is yet another example of the potential of combining
201 SCR methods and non-invasive sampling techniques to estimate abundance and density for
202 elusive and wide-ranging species, like large carnivores [13–18].
203 When examining densities across study areas in the French Jura mountain, we found
204 spatial variation between the three counties, with Doubs area having the lowest densities, Ain
205 the highest densities, and Jura intermediate densities. Our density estimates were of similar
206 magnitude to other lynx populations in Europe: 1.47 and 1.38 lynx / 100 km2 in the
207 Northwestern Swiss Alps [13], 0.58 (Štiavnica mountains) and 0.81 individuals / 100 km2
208 (Velká Fatra National Park) in Slovakia [37] and 0.9 individuals / 100 km2 in the Bavarian
209 Forest National Park in Germany [38].
210 While [13] and [37] used SCR models, [38] used standard capture-recapture models
211 with HMMDM to estimate densities, which makes them difficult to compare [39]. Indeed, in
212 other carnivore studies, the use of HMMDM also produced similar density estimates to SCR
213 models [13], while in others, including ours, the SCR estimates were closer to the MMDM
214 estimates [2] or intermediate between the MMDM and HMMDM estimates [40]. When
215 looking at reference values for densities across the distribution range of the species, it may be
216 biologically meaningful to use the MMDM density estimate as a reference as it covers the
217 whole potential of animal movements. On the other hand, because SCR models make space
218 explicit whereas standard model-based densities are sensitive to the definition of the effective
219 sampling area, we recommend the use of SCR models to estimate lynx densities. Our lynx
220 density estimates might suffer from potential sources of bias that need to be discussed. First,
221 the period of sampling is important to account for when setting up camera trap surveys [41].
222 We conducted our survey outside the dispersal period, during the lynx mating season
223 (February-March mostly). We did so to avoid capturing transient individuals and to increase
224 detectability because of high lynx activity and relatively reduced human activities [31]. bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
225 However, some individuals might have moved in and out of the study areas, especially males
226 who cover greater distances during the mating season. Whereas the presence of non-resident
227 individuals can affect the calculation of (H)MMDM, and in turn density estimated with
228 standard capture-recapture models, SCR density estimates were found to be robust to the
229 presence of transient individuals [42]. Second, males have larger home ranges than females
230 [13], which leads to heterogeneity in the SCR model parameter estimates. Because there were
231 too few males and females identified and lots of individuals with unknown sex, sex-specific
232 SCR analyses [43] produced unreliable abundance and density estimates (results not shown).
233 If detection heterogeneity is ignored in capture-recapture models, abundance is
234 underestimated [44], therefore our density estimates are probably biased low and should be
235 considered as a conservative metric. The determination of sex could be improved by i)
236 combining the photographic surveys with genetic surveys, ii) conducting deterministic
237 surveys over several years (e.g., [13]), iii) conducting an opportunistic camera trapping survey
238 all over the years and setting camera trap at fresh lynx kills, iv) setting infrared flash camera
239 traps capable of taking burst of images in rapid sequence at marking sites regularly used by
240 the lynx (e.g., [45]). Last, we did not detect any individuals in the Vosges mountain, even
241 though the sampling effort was similar to that implemented in the Jura mountain (Table 1).
242 This finding is likely to be representative of the current critical situation of the lynx in the
243 Vosges mountain.
244 We envision several perspectives to our work. First, while density estimates are of
245 primary interest for conservation, understanding the mechanisms underlying trends in
246 abundance is required to make sound conservation decisions [1]. SCR models have been
247 extended to open populations [46] and can be used to estimate demographic parameters
248 (survival, reproduction) of large carnivores [47]. Unfortunately, because of logistic
249 constraints, we could not sample the same areas over several years, which precludes a bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
250 standard application of these models. A solution may lie in the combination of the data we
251 collected through systematic camera-trap surveys with additional data in the SCR framework,
252 such as occupancy data [48] or opportunistic camera-trap data [49]. Second, in addition to
253 traffic-induced mortality and conflicts with human activities, the expansion of lynx
254 populations is limited by habitat fragmentation [50], hence the need to assess connectivity
255 with other populations [51]. SCR models can be used to quantify landscape connectivity by
256 replacing the Euclidean distance between camera traps and home range centers by the
257 least‐cost path [52,53]. For lynx, this will require setting up traps across a gradient of habitat
258 types, not only forested habitats, so that resistance to movement can be estimated.
259 In conclusion, our lynx density estimates for the French Jura mountain complement
260 nicely the estimates recently provided for the Northwestern Swiss Alps [13]. The use of
261 camera-trapping coupled with SCR models in both France and Switzerland was the result of a
262 cooperation between the two countries with the perspective of a transboundary monitoring
263 [54,55]. This approach would prove useful to accurately estimate densities in other areas
264 where habitats and prey availability might differ, and overall lynx detectability varies. Also,
265 collecting and adding movement data from GPS-collared lynx would be useful [49,56] to try
266 and infer the connections between subpopulations.
267 The case can be made for monitoring the return of the lynx in the French Alps. Indeed,
268 small-scale camera-trapping surveys and opportunistic observations are currently active and
269 producing signs of lynx presence. However, the lack of a coordinated and intensive sampling
270 effort prevents us from being able to estimate abundance and density and inferring trends.
271 In contrast, the situation in the Vosges mountain is alarming with no individuals
272 detected over the study period. Because the Vosges mountain are located between the French
273 Jura mountain and the Palatinate Forest in Germany where a reintroduction program is
274 ongoing (program LIFE13 NAT/DE/000755), the lynx colonization in the Vosges mountain bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
275 remains possible both by the north and the south. Incidentally, two cases of lynx dispersal in
276 the Vosges mountain from neighboring mountains have been recently observed ([57];
277 program LIFE13 NAT/DE/000755). To ensure the detection of lynx in the Vosges mountain,
278 we recommend reinforcing collaborative monitoring by involving all field stakeholders and
279 enhancing communication on the species signs of presence.
280 In this context, obtaining accurate and comparable lynx densities will be crucial to
281 closely monitor population trends at the national scale and inform management policies for
282 the effective conservation of the Eurasian lynx in France.
283
284 Acknowledgments
285 We thank the staff from the French National Game and Wildlife Agency (ONCFS), the
286 CROC, the Forest National Agency, the “Directions Départementales des Territoires”, the
287 “Fédérations Départementales des Chasseurs” the Regional Natural Parks, the environmental
288 protection associations and all the volunteers from the “Réseau Loup Lynx” who collected the
289 photographs during the camera-trapping session. OG was funded by CNRS and the “Mission
290 pour l’Interdisciplinarité” through the “Osez l’Interdisciplinarité” initiative. CD, EM and SG
291 were funded by ONCFS. CEFE and CROC were funded by CIL&B, MTES (ITTECOP) and
292 FRB through the research program ERC-Lynx. CROC was funded by the European Union
293 within the framework of the Operational Program FEDER-FSE “Lorraine et Massif des
294 Vosges 2014–2020”, the “Commissariat à l’Aménagement du Massif des Vosges” for the
295 FNADT (“Fonds National d'Aménagement et de Développement du Territoire”), the DREAL
296 Grand Est (“Direction Régionale pour l’Environnement, l’Aménagement et le Logement”),
297 the “Région Grand Est”, the “Zoo d’Amnéville”, the “Fondation d’entreprise UEM”, the
298 “Fondation Nature & Découvertes”, the “Fondation Le Pal Nature”, and the Chasseur bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
299 d’images magazine. This study could not have been conducted without authorizations from
300 and agreements with municipalities, environmental managers and owners.
301
302 Author contributions
303 OG wrote the paper and all co-authors commented on the manuscript. OG and SG analyzed
304 the data. AL, CD, EG, EM and SG coordinated the study designs, the data collection and
305 interpretation, with help from FZ for setting the experimental design in the Jura mountain.
306
307 Data accessibility
308 The Eurasian lynx is an endangered species with high conservation stakes. Interactions with
309 human activities are problematic and lead to poaching and anthropogenic pressures. Providing
310 accurate information on lynx locations can be detrimental to the conservation status of the
311 species. As a consequence, the original data could not be shared.
312
313 References
314 1. Williams BK, Nichols JD, Conroy MJ. Analysis and management of animal
315 populations. San Diego: Academic Press; 2002.
316 2. Obbard ME, Howe EJ, Kyle CJ. Empirical comparison of density estimators for large
317 carnivores. Journal of Applied Ecology. 2010;47: 76–84. doi:10.1111/j.1365-
318 2664.2009.01758.x
319 3. Kelly M, Betsch J, Wultsch C, Mesa B, Mills L. Noninvasive Sampling for
320 Carnivores. Carnivore Ecology and Conservation: A Handbook of Techniques. Oxford
321 University Press. 2012. pp. 47–69.
322 4. Royle JA, Chandler RB, Sollmann R, Gardner B. Spatial Capture-Recapture [Internet].
323 New York, USA: Academic Press, New York; 2014. doi:10.1016/C2012-0-01222-7 bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
324 5. Otis DL, Burnham KP, White GC, Anderson DR. Statistical Inference from Capture
325 Data on Closed Animal Populations. Wildlife Monographs. The Wildlife Society; 1978. pp.
326 3–135. doi:10.2307/2287873
327 6. Mumma MA, Zieminski C, Fuller TK, Mahoney SP, Waits LP. Evaluating
328 noninvasive genetic sampling techniques to estimate large carnivore abundance. Molecular
329 Ecology Resources. 2015;15: 1133–1144. doi:10.1111/1755-0998.12390
330 7. Gerber BD, Ivan JS, Burnham KP. Estimating the abundance of rare and elusive
331 carnivores from photographic-sampling data when the population size is very small.
332 Population Ecology. 2014;56: 463–470. doi:10.1007/s10144-014-0431-8
333 8. White GC, Anderson DR, Burnham KP, Otis DL. Capture-recapture and removal
334 methods for sampling closed populations. Los Alamos, NM: Los Alamos National
335 Laboratory; 1982 p. 14. Available: http://pubs.er.usgs.gov/publication/70120390
336 9. Royle JA, Young KV. A hierarchical model for spatial capture-recapture data.
337 Ecology. 2008;89: 2281–2289. doi:10.1890/07-0601.1
338 10. Borchers DL, Efford MG. Spatially Explicit Maximum Likelihood Methods for
339 Capture–Recapture Studies. Biometrics. 2008;64: 377–385. doi:10.1111/j.1541-
340 0420.2007.00927.x
341 11. Efford M. Density estimation in live-trapping studies. Oikos. 2004;106: 598–610.
342 doi:10.1111/j.0030-1299.2004.13043.x
343 12. Borchers D. A non-technical overview of spatially explicit capture-recapture models.
344 Journal of Ornithology. 2012;152: 435–444.
345 13. Pesenti E, Zimmermann F. Density estimations of the Eurasian lynx ( Lynx lynx ) in
346 the Swiss Alps. Journal of Mammalogy. 2013;94: 73–81. doi:10.1644/11-MAMM-A-322.1
347 14. Stetz JB, Mitchell MS, Kendall KC. Using spatially-explicit capture-recapture models
348 to explain variation in seasonal density patterns of sympatric ursids. Ecography. 2018; bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
349 doi:10.1111/ecog.03556
350 15. López-Bao JV, Godinho R, Pacheco C, Lema FJ, García E, Llaneza L, et al. Toward
351 reliable population estimates of wolves by combining spatial capture-recapture models and
352 non-invasive DNA monitoring. Scientific Reports. 2018;8. doi:10.1038/s41598-018-20675-9
353 16. Goldberg JF, Tempa T, Norbu N, Hebblewhite M, Mills LS, Wangchuk TR, et al.
354 Examining Temporal Sample Scale and Model Choice with Spatial Capture-Recapture
355 Models in the Common Leopard Panthera pardus. Aegerter CM, editor. PLOS ONE. 2015;10:
356 e0140757. doi:10.1371/journal.pone.0140757
357 17. Broekhuis F, Gopalaswamy AM. Counting Cats: Spatially Explicit Population
358 Estimates of Cheetah (Acinonyx jubatus) Using Unstructured Sampling Data. Festa-Bianchet
359 M, editor. PLOS ONE. 2016;11: e0153875. doi:10.1371/journal.pone.0153875
360 18. Alexander JS, Gopalaswamy AM, Shi K, Riordan P. Face Value: Towards Robust
361 Estimates of Snow Leopard Densities. Margalida A, editor. PLOS ONE. 2015;10: e0134815.
362 doi:10.1371/journal.pone.0134815
363 19. Chapron G, Kaczensky P, Linnell JDC, von Arx M, Huber D, Andren H, et al.
364 Recovery of large carnivores in Europe’s modern human-dominated landscapes. Science.
365 2014;346: 1517–1519.
366 20. Breitenmoser U. Large predators in the Alps: The fall and rise of man’s competitors.
367 Biological Conservation. 1998;83: 279–289. doi:10.1016/S0006-3207(97)00084-0
368 21. Vandel J-M, Stahl P. Distribution trend of the Eurasian lynx Lynx lynx populations in
369 France. Mammalia. 2005;69. doi:10.1515/mamm.2005.013
370 22. Breitenmoser U, Breitenmoser-Würsten C, Capt S. Re-introduction and present status
371 of the lynx ( Lynx Lynx ) in Switzerland. Hystrix It J Mamm. 1998;10. doi:10.4404/hystrix-
372 10.1-4118
373 23. Vandel J-M, Stahl P, Herrenschmidt V, Marboutin E. Reintroduction of the lynx into bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
374 the Vosges mountain massif: From animal survival and movements to population
375 development. Biological Conservation. 2006;131: 370–385.
376 doi:10.1016/j.biocon.2006.02.012
377 24. Marboutin E, Duchamp C, Moris P, Briaudet P-E, Léonard Y, Catusse M. Survey of
378 the Lynx distribution in the French Alps : 2005-2009 update. Acta biologica Slovenica.
379 2012: 29–34.
380 25. Laurent A, Léger F, Briaudet PE, Léonard Y, Bataille A, Goujon G. Evolution récente
381 (2008-2010) de la population de Lynx en France. Faune Sauvage. 2012;294: 38–39.
382 26. Breitenmoser U, Breitenmoser-Würsten C, Capt S, Molinari-Jobin A, Molinari P,
383 Zimmermann F. Conservation of the lynx Lynx lynx in the Swiss Jura Mountains. Wildlife
384 Biology. 2007;13: 340–355. doi:10.2981/0909-6396(2007)13[340:COTLLL]2.0.CO;2
385 27. DREAL Grand Est. Diagnostic partagé du massif des Vosges pour une vision
386 transversale du territoire et de ses enjeux. 2018.
387 28. Zimmermann F, Breitenmoser-Würsten C, Molinari-Jobin A, Breitenmoser U.
388 Optimizing the size of the area surveyed for monitoring a Eurasian lynx (Lynx lynx)
389 population in the Swiss Alps by means of photographic capture-recapture. Integr Zool.
390 2013;8: 232–243. doi:10.1111/1749-4877.12017
391 29. Breitenmoser-Würsten C, Zimmermann F, Stahl P, Vandel J-M,
392 Molinari-Jobin A, Molinari P, Capt S, Breitenmoser U. Spatial and Social stability of a
393 Eurasian lynx Lynx lynx population: an assessment of 10 years of observation in the Jura
394 Mountains. Wildlife Biology. 2007;13: 365–380. doi:10.2981/0909-
395 6396(2007)13[365:SASSOA]2.0.CO;2
396 30. Blanc L, Marboutin E, Gatti S, Gimenez O. Abundance of rare and elusive species:
397 Empirical investigation of closed versus spatially explicit capture-recapture models with lynx
398 as a case study. The Journal of Wildlife Management. 2013;77: 372–378. bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
399 doi:10.1002/jwmg.453
400 31. Zimmermann F, Foresti D. Capture-recapture methods for density estimation. Camera
401 Trapping for Wildlife Research. Pelagic Publishing. Exeter, UK; 2016. pp. 95–141.
402 32. Zimmermann F, Breitenmoser‐Würsten C, Breitenmoser U. Natal dispersal of
403 Eurasian lynx (Lynx lynx) in Switzerland. Journal of Zoology. 2005;267: 381–395.
404 doi:10.1017/S0952836905007545
405 33. Rocha DG da, Sollmann R, Ramalho EE, Ilha R, Tan CKW. Ocelot (Leopardus
406 pardalis) Density in Central Amazonia. Hagen CA, editor. PLOS ONE. 2016;11: e0154624.
407 doi:10.1371/journal.pone.0154624
408 34. Sutherland C, Muñoz DJ, Miller DAW, Grant EHC. Spatial Capture–Recapture: A
409 Promising Method for Analyzing Data Collected Using Artificial Cover Objects.
410 Herpetologica. 2016;72: 6–12. doi:10.1655/HERPETOLOGICA-D-15-00027
411 35. Sutherland C, Royle JA, Linden DW. oSCR: a spatial capture–recapture R package for
412 inference about spatial ecological processes. Ecography. 0. doi:10.1111/ecog.04551
413 36. Baillargeon S, Rivest L-P. Rcapture: Loglinear Models for Capture-Recapture in R.
414 Journal Of Statistical Software. 2007;19: 1–31.
415 37. Kubala J, Smolko P, Zimmermann F, Rigg R, Tám B, Iľko T, et al. Robust monitoring
416 of the Eurasian lynx Lynx lynx in the Slovak Carpathians reveals lower numbers than
417 officially reported. Oryx. 2017; 1–9. doi:10.1017/S003060531700076X
418 38. Weingarth K, Heibl C, Knauer F, Zimmermann F, Bufka L, Heurich M. First
419 estimation of Eurasian lynx (Lynx lynx) abundance and density using digital cameras and
420 capture–recapture techniques in a German national park. Animal Biodiversity and
421 Conservation. 2012; 11.
422 39. Gerber BD, Karpanty SM, Kelly MJ. Evaluating the potential biases in carnivore
423 capture–recapture studies associated with the use of lure and varying density estimation bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
424 techniques using photographic-sampling data of the Malagasy civet. Popul Ecol. 2012;54: 43–
425 54. doi:10.1007/s10144-011-0276-3
426 40. Reppucci J, Gardner B, Lucherini M. Estimating detection and density of the Andean
427 cat in the high Andes. Journal of Mammalogy. 2011;92: 140–147. doi:10.1644/10-MAMM-
428 A-053.1
429 41. Weingarth K, Zeppenfeld T, Heibl C, Heurich M, Bufka L, Daniszová K, et al. Hide
430 and seek: extended camera-trap session lengths and autumn provide best parameters for
431 estimating lynx densities in mountainous areas. Biodiversity and Conservation. 2015;24:
432 2935–2952. doi:10.1007/s10531-015-0986-5
433 42. Royle JA, Fuller AK, Sutherland C. Spatial capture–recapture models allowing
434 Markovian transience or dispersal. Population Ecology. 2016;58: 53–62. doi:10.1007/s10144-
435 015-0524-z
436 43. Sollmann R, Furtado MM, Gardner B, Hofer H, Jácomo ATA, Tôrres NM, et al.
437 Improving density estimates for elusive carnivores: Accounting for sex-specific detection and
438 movements using spatial capture–recapture models for jaguars in central Brazil. Biological
439 Conservation. 2011;144: 1017–1024. doi:10.1016/j.biocon.2010.12.011
440 44. Cubaynes S, Pradel R, Choquet R, Duchamp C, Gaillard J-M, Lebreton J-D, et al.
441 Importance of accounting for detection heterogeneity when estimating abundance: the case of
442 French wolves. Conservation Biology. 2010;24: 621–6. doi:10.1111/j.1523-
443 1739.2009.01431.x
444 45. Vogt K, Zimmermann F, Kölliker M, Breitenmoser U. Scent-marking behaviour and
445 social dynamics in a wild population of Eurasian lynx Lynx lynx. Behavioural Processes.
446 2014;106: 98–106. doi:10.1016/j.beproc.2014.04.017
447 46. Gardner B, Reppucci J, Lucherini M, Royle JA. Spatially explicit inference for open
448 populations: Estimating demographic parameters from camera-trap studies. Ecology. 2010;91: bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
449 3376–3383.
450 47. Whittington J, Sawaya MA. A Comparison of Grizzly Bear Demographic Parameters
451 Estimated from Non-Spatial and Spatial Open Population Capture-Recapture Models. Boyce
452 MS, editor. PLOS ONE. 2015;10: e0134446. doi:10.1371/journal.pone.0134446
453 48. Blanc L, Marboutin E, Gatti S, Zimmermann F, Gimenez O. Improving abundance
454 estimation by combining capture-recapture and occupancy data: Example with a large
455 carnivore. Journal of Applied Ecology. 2014;51: 1733–1739.
456 49. Tenan S, Pedrini P, Bragalanti N, Groff C, Sutherland C. Data integration for
457 inference about spatial processes: A model-based approach to test and account for data
458 inconsistency. Rocchini D, editor. PLOS ONE. 2017;12: e0185588.
459 doi:10.1371/journal.pone.0185588
460 50. Kramer-Schadt S, Revilla E, Wiegand T, Breitenmoser U. Fragmented landscapes,
461 road mortality and patch connectivity: Modelling influences on the dispersal of Eurasian lynx.
462 Journal of Applied Ecology. 2004;41: 711–723. doi:10.1111/j.0021-8901.2004.00933.x
463 51. Zimmermann F, Breitenmoser U. Potential distribution and population size of the
464 Eurasian lynx Lynx lynx in the jura Mountains and possible corridors to adjacent ranges.
465 Wildlife Biology. 2007;13: 406–416. doi:10.2981/0909-
466 6396(2007)13[406:PDAPSO]2.0.CO;2
467 52. Royle JA, Chandler RB, Gazenski KD, Graves TA. Spatial capture–recapture models
468 for jointly estimating population density and landscape connectivity. Ecology. 2013;94: 287–
469 294. doi:10.1890/12-0413.1
470 53. Sutherland C, Fuller AK, Royle JA. Modelling non-Euclidean movement and
471 landscape connectivity in highly structured ecological networks. Methods in Ecology and
472 Evolution. 2015;6: 169–177. doi:10.1111/2041-210X.12316
473 54. Gervasi V, Brøseth H, Gimenez O, Nilsen EB, Odden J, Flagstad ., et al. Sharing data bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
474 improves monitoring of trans-boundary populations: The case of wolverines in central
475 Scandinavia. Wildlife Biology. 2016;22: 95–106. doi:10.2981/wlb.00142
476 55. Vitkalova AV, Feng L, Rybin AN, Gerber BD, Miquelle DG, Wang T, et al.
477 Transboundary cooperation improves endangered species monitoring and conservation
478 actions: A case study of the global population of Amur leopards. Conservation Letters. 2018;
479 e12574. doi:10.1111/conl.12574
480 56. Linden DW, Sirén APK, Pekins PJ. Integrating telemetry data into spatial capture–
481 recapture modifies inferences on multi-scale resource selection. Ecosphere. 2018;9: e02203.
482 doi:10.1002/ecs2.2203
483 57. Hurstel A, Laurent A. Première preuve de dispersion du Lynx d’Eurasie (Lynx lynx)
484 du Jura vers les Vosges. Ciconia. 2016: 1–6.
485 58. Efford M. secr: Spatially explicit capture-recapture models. R package version 3.2.1.
486 [Internet]. 2019. Available: https://CRAN.R-project.org/package=secr
487
488 bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
489 Figures
490
491 Figure 1: Map of the study area in the French Jura and Vosges mountains. The study
492 area encompassed seven counties (Ain, Jura and Doubs in the Jura mountain and Vosges,
493 Haut-Rhin, Bas-Rhin and Moselle in the Vosges mountain) that were monitored through 418
494 camera trapping sites (298 in the Jura mountain and 115 in the Vosges mountain; two camera
495 traps were set per site), each within a 2.7 x 2.7 km cell. The inset map represents the French
496 counties (grey borders), the counties that were considered in the study (black borders), the
497 Jura mountain (green shaded area) and the Vosges mountain (red shaded area).
498
499
500
501 Figure 2: Lynx (Lynx lynx) density maps in the French Jura mountain. The density scale
502 is in lynx per 2.25 km2 (pixel resolution is 1500m x 1500m). We obtained the estimated
503 abundance in each map by summing up the densities in each pixel altogether. Yellow is for
504 low densities, green for medium densities and blue for high densities; the density scales are
505 specific to each map. Note that the interpretation of these plots as density maps is subject to
506 caution (see the vignette “secr-densitysurface” of the SECR R package [58]).
507
508 509 510 certified bypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmadeavailableunder bioRxiv preprint
Table 1: Main characteristics and results of the lynx camera-trap survey carried out in a) the French Jura mountain and b) the French
Vosges mountain. doi:
https://doi.org/10.1101/600015 a. Year/County 2011/Doubs 2011/Jura 2012/Jura & Doubs 2013/Doubs 2013/Ain & Jura 2014/Ain 2015/Ain Period of trap activity Jan-Apr Feb-Apr Feb-Apr Feb-Apr Feb-Apr Feb-Apr Feb-May Number of active camera traps 48 66 148 44 142 118 30 Number of trapping days (average/area) 63 59 69 63 58 59 99 Number of capture occasionsi 15 15 17 14 13 13 21 a CC-BY-NC-ND 4.0Internationallicense ;
Number of detections 22 42 130 25 117 158 38 this versionpostedAugust4,2019. Number of detected individuals 4 9 21 6 19 23 10 Number of females, unknown, males 1, 1, 2 1, 7, 1 2, 14, 5 1, 4, 1 2, 13, 4 4, 16, 3 2, 8, 0 Number of detections / ind: mean, min, max 3, 2, 4 2.8, 1, 6 2.5, 1, 10 2.7, 1, 6 3.6, 1, 11 3.3, 1, 9 2.2, 1, 5 b. Year/County 2013/Haut-Rhin & Vosges 2014/Bas-Rhin & Moselle 2015/Bas-Rhin & Moselle 2016/Bas & Haut-Rhin Period of trap activity Dec-Jan Feb-Apr Feb-Apr Feb-Apr The copyrightholderforthispreprint(whichwasnot
Number of active traps 60 50 60 60 . Number of trapping days (average/area) 52 59 57 59 Number of detections 0 0 0 0 iA capture occasion is defined as 5 successive trap days.
certified bypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmadeavailableunder bioRxiv preprint
Table 2: Lynx abundance and density estimates obtained from spatial and non-spatial capture-recapture analyses of camera-trapping doi: 2 data collected in the French Jura mountain. Densities are provided in number of lynx per 100 km . For 2011 and 2013, parameters of the https://doi.org/10.1101/600015
i spatial capture-recapture model (p0 and σ) are common to both areas in each year. Acronyms are defined in the footnote .
Year/County 2011/Doubs 2011/Jura 2012/Jura-Doubs 2013/Doubs 2013/Ain-Jura 2014-Ain 2015-Ain a CC-BY-NC-ND 4.0Internationallicense
SCR abundance (SE) 5 (0.1) 12 (0.1) 29 (0.2) 7 (0.1) 21 (0.1) 29 (0.2) 12 (0.1) ; this versionpostedAugust4,2019. SCR density (SE) 0.24 (0.02) 0.44 (0.02) 0.67 (0.02) 0.36 (0.02) 0.54 (0.02) 0.91 (0.03) 0.64 (0.03) p0 logit scale (SE) -2.94 (0.24) -2.01 (0.20) -2.57 (0.20) -2.34 (0.19) -3.01 (0.42)
σ log scale (SE) 8.89 (0.14) 8.54 (0.08) 8.95 (0.06) 8.80 (0.07) 8.97 (0.19)
M0 abundance (SE) 4 (0.7) 9 (0.7) 21 (0.6) 6 (0.3) 19 (0.8) 23 (0.7) 11 (1.2)
Mh abundance (SE) 5 (1.7) 10 (1.8) 25 (2.8) 7 (1.2) 25 (4.1) 28 (3.6) 11 (1.2) The copyrightholderforthispreprint(whichwasnot . MMDM (km) 9.1 16.2 8.9 9.1 18.2 13.6 12.1
ETA with MMDM (km2) 1991 2930 3089 1171 4954 2936 1549
M0 density MMDM (SE) 0.31 (0.05) 0.31 (0.02) 0.68 (0.02) 0.51 (0.02) 0.38 (0.02) 0.78 (0.02) 0.71 (0.08)
Mh density MMDM (SE) 0.39 (0.13) 0.34 (0.06) 0.81 (0.09) 0.60 (0.10) 0.50 (0.08) 0.95 (0.12) 0.70 (0.08)
ETA with HMMDM (km2) 697 1491 2111 659 2673 1668 753 certified bypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmadeavailableunder bioRxiv preprint
M0 density HMMDM (SE) 0.57 (0.10) 0.60 (0.05) 0.99 (0.03) 0.91 (0.05) 0.71 (0.03) 1.38 (0.04) 1.46 (0.16) doi:
Mh density HMMDM (SE) 0.72 (0.24) 0.67 (0.12) 1.18 (0.13) 1.06 (0.18) 0.93 (0.15) 1.68 (0.21) 1.43 (0.16) https://doi.org/10.1101/600015
iAcronyms used: ETA is for Effective Trapping Area, MMDM for Mean Maximum Distance Moved, HMMDM for Half Mean Maximum
Distance Moved, SCR for spatial capture-recapture, M0 for the (non-spatial) capture-recapture model with homogeneous detection probability,
Mh for the (non-spatial) capture-recapture model with heterogeneous detection probability and SE for standard error. a
CC-BY-NC-ND 4.0Internationallicense ; this versionpostedAugust4,2019.
The copyrightholderforthispreprint(whichwasnot . bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Figure 1 bioRxiv preprint doi: https://doi.org/10.1101/600015; this version posted August 4, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Figure 2